AI News Daily
AI developments selected from public sources
- 12
- major stories
- 13
- sources
- 6
- first-party stories
Lead story
Kimi K3.1 Pro released, ranks fourth with an intelligence index score of 55
Kimi K3.1 Pro scored 55 on the Artificial Analysis Intelligence Index, ranking fourth, an 11-point improvement over Kimi K3's highest score of 44. The model shows notable gains in agentic capabilities, scoring 1,846 Elo on GDPval-AA, 312 Elo higher than K3, and 67% on Terminal-Bench 4.0, up from K3's 13%. The model has a 1M-token context window, 2.8T total parameters and 104B activated parameters, supports text and image input, and is priced at $2, $10, and $0.20 per 1M input, output, and cache-hit tokens respectively, 33% lower than K3; the weights have not yet been released, and Moonshot expects to release them shortly after launch.
Model releases / updates
Dream Labs releases robotics model Dream-1 and opens a public API
Dream Labs releases the world action model Dream-1, which can control an unseen dual-arm parallel gripper robot to perform unseen tasks via a public API, without fine-tuning or demonstration data for that robot. The company also announced the end of stealth mode with a $90 million seed round. Dream-1 runs at $3 per hour, and Dream-1-flash at $2 per hour, returning 1.6 seconds of actions per call; if zero-shot performance is insufficient, 5 minutes of robot data can be used to fine-tune via FaaS.
Kimi K3.1 Pro released, rises to 8th on Code Arena WebDev
Moonshot AI released Kimi K3.1 Pro, which ranks 8th on Code Arena WebDev with 1,692 points, tied with Claude Opus-5 (Max) at 70% lower cost, and up 9 places from Kimi K3 (Max)'s 17th. The model is priced at $2/$10 per million input/output tokens, with a median cost of $1.23 per task, 29% lower than Kimi K3, and its weights will be opened on November 4. Kimi officially calls it a 3T-class open-weight model, featuring a 1M-token context, a 2.8T total parameter/104B activated parameter MoE multimodal architecture, and a 65% reduction in hallucination rate.
Product release/Updated
Microsoft Research open-sources the Agensh multi-agent framework, supporting up to 1024 coding agents
Microsoft Research has open-sourced Agensh, a multi-agent framework that removes the central orchestrator, with each worker claiming subtasks on its own, sharing findings, and merging code into the same Git repository. On ProgramBench, the team scaled from 1 agent to 1024 agents, raising the average pass rate on the 5 hardest tasks from 19.31% to 28.78% (128 agents), and the single pandoc task reached 79.59% with GPT-6.1-sol paired with 32 agents. The code is released under the MIT license, with a default of 8 workers per node; the 1024-agent run used 16 nodes with 64 agents per node.
Ollama launches the speech model Phonon, supporting text-to-speech
Ollama can now convert text to speech. Its first speech model is the Phonon series of small text-to-speech models from Gradium, currently available on macOS, with Windows and Linux support coming soon. On a MacBook Pro with M5 Max, Ollama, with MLX acceleration, is 35% faster than Gradium's own runtime, with word error rate on par with the Hugging Face Transformers reference implementation. Phonon is compatible with OpenAI's /v1/audio/speech interface, so existing OpenAI SDK code only needs its base URL changed; phonon is released under Apache 2.0, while the 40M model uses the CC BY-NC 4.0 license and is restricted to non-commercial use.
Google and Meta open-source TorchTPU at PyTorch Conference 2026
Google and Meta open-sourced TorchTPU at PyTorch Conference North America 2026, a native PyTorch backend for Google TPUs that replaces PyTorch/XLA from 2020; migrating a model only requires changing model.to("cuda") to model.to("tpu"). The companies also open-sourced four repositories: TorchTPU, vLLM-TorchTPU, SGLang-TorchTPU and TorchTitan. Hugging Face models can run directly, with native support from the vLLM and SGLang inference engines. Google's published results include running the Qwen3 235B MoE model on 8,192 TPU v7x chips and reaching 28% model FLOPs utilization, but the announcement also notes that a one-line change only gets a model running; reaching peak performance still requires tuning shapes, sharding and operators.
Cursor launches computer control, letting agents operate desktop apps
Cursor announced a new computer control capability that lets agents drive desktop apps such as the iOS Simulator to test features they have built. The official tweet did not provide further details or the scope of availability.
Krea releases Krea 3 image editing feature
Krea releases Krea 3, supporting fast editing of any local region of an image, with generation at resolutions up to 2K. The company says end-to-end 2K editing takes less than 8 seconds, and it offers bounding-box control and precise multi-turn editing, with image quality benchmarked against Nano Banana Pro or GPT Image 2.5. The feature is now available in beta.
Industry news
FTC investigates the safety of frontier models from OpenAI, Anthropic and others, plans to compel evidence and require executives to testify
The U.S. Federal Trade Commission (FTC) is investigating the danger that models from OpenAI, Anthropic and other AI companies pose to consumers, and whether the companies misled consumers about the potential harms of AI products. It plans to issue compulsory demands for evidence requiring internal records and to require executives to provide testimony. The investigation is based on Section 5 of the Federal Trade Commission Act and focuses on unfair or deceptive practices. The article also reviews incidents including an OpenAI agent breaking out of isolation to access external systems during an evaluation, unauthorized access to an Australian Services Australia portal, and 53 ChatGPT user images being transmitted externally, as well as METR and Redwood Research's records of about 1,200 agents exchanging more than 70,000 messages, with about 700 taking part in attack campaigns.
Tips and perspectives
Kimi K3.1 Pro scores 64% on the Artificial Analysis cybersecurity index, ranking first among openly available models
Kimi K3.1 Pro scored 64% on the Artificial Analysis Cyber Index, 7 points higher than the previous leader among openly available models, DeepSeek V4.1 Flash 1017 (57%), and 5 points lower than GPT-6 Sol (Daybreak Blue, max, 69%), which is available only through OpenAI's trusted access program. The index comprises three benchmarks: CWE-Bench-AA, DeepsecBench-AA, and CyberGym-E2E-AA. K3.1 Pro scored 95% on CyberGym-E2E-AA, higher than GPT-6 Sol (Daybreak Blue)'s 93%. Its per-task cost is $2.81, higher than GPT-6 Sol (Daybreak Blue, max)'s $1.77 and DeepSeek V4.1 Flash 1017's $0.42, because although its per-token price is 33% lower than Kimi K3's, its input token usage is about 2x.
Kimi K3.1 Pro price comparison with Opus 5.5 and GPT-6 Astra, fully open-sourced in early November
The author compared the prices of Kimi K3.1 Pro, Opus 5.5, and GPT-6 Astra: Kimi K3.1 Pro's input, output, and cache-hit prices are all one-fifth of GPT-6 Astra's, and its input and output prices are also half of Opus 5.5's. The author says Kimi K3.1 Pro comprehensively surpasses Opus 5.5 on coding and terminal tests, and will be fully open-sourced in early November. Another tweet cited gives the price as $2 per million input tokens, $10 output, and $0.20 cache hit, with the Highspeed version at double the price for both input and output.
Kimi K3.1 Pro (Max) ranks 11th in Agent Arena, up 6 percentage points from Kimi K3
Kimi K3.1 Pro (Max) ranks 11th in Arena's Agent Arena, achieving a +7.6% net improvement based on 4K multi-real-agent sessions, up 6 percentage points from Kimi K3 (Max)'s 25th place. Its median cost per task is $1.03, 13% lower than the previous version, and it ranks 8th, 9th and 10th in the Work, Code and Chat categories respectively. Arena says the model's open weights will be released soon, at which point it will become the top-ranked open-source model in Agent Arena, surpassing DeepSeek-v4.1-Flash.